Abstract
In visual search, attentional selection is guided by attentional templates (target representations held in visual working memory). In colour search, it was found that multiple target templates can be activated to guide attentional selection in parallel. However, it remains unclear whether multiple templates can also be activated simultaneously when they specify combinations of features from different feature dimensions (colour and shape) as needed during conjunction search. To answer this question, we measured N2pc components of the event-related potential under low and high load conditions in colour, shape, and colour/shape conjunction search. Search displays in the three task conditions were physically identical. They each contained four outline objects of different colour-shape combinations. In different blocks, each search display was preceded by a cue display indicating the relevant one or two target colour(s), shape(s), or colour-shape combination(s), for the upcoming search, respectively. Participants task was to find the target object that matched (one of) the cue(s) and report the orientation of a line shown within the object (horizontal/vertical). N2pc components measured in the colour task mirrored previous findings. They were slightly delayed and attenuated in high- versus low-load trials, reflecting mutual inhibition of two simultaneously activated colour templates. The same N2pc pattern was found in the shape search. Importantly, the relative load costs (latency and amplitude differences in high- minus low-load trials) were comparable between the colour and shape tasks. This suggests that multiple shape templates, just like colour templates, can be activated in parallel if multiple shapes are task relevant. In contrast, in the conjunction search, N2pc amplitudes were attenuated in the high-load relative to the low-load condition by approximately 50%, indicating that only a single conjunctive template could be activated at a time. This suggests that attentional templates in conjunction search hold separate features rather than a combined object representation.